RepoRadar
Job search tool that scans GitHub to find companies building with specific tech stacks. Instead of applying to LinkedIn ghost jobs, search for companies that use the same tools you do by reading their dependency files, then reach out directly. Scores prospects 0–100 based on stack match, AI tool signals, and activity. Automatically checks Greenhouse, Lever, Ashby, and Workable for open roles.
Project Overview
RepoRadar flips job searching on its head. Instead of 'find jobs, then figure out if the company matches,' it's 'find companies that build like you build, then reach out directly.' It scans GitHub for requirements.txt, package.json, CLAUDE.md and other dependency files, detects tech stacks across 200+ known packages, scores prospects, probes ATS job boards, and generates AI-powered personalized outreach.
The Challenge
Job searching on LinkedIn is broken. Posts are stale, ghost jobs are everywhere, and applications disappear into a black hole. But there's a signal nobody's using: GitHub.
The Solution
Built a tool that scans GitHub to find companies using the same tech stack you do, scores them 0-100, checks their ATS job boards for open roles, and generates personalized outreach using Claude AI.
Technology Stack
Backend
Frontend
Ai
Integrations
Deployment
Key Features
GitHub code search scanning requirements.txt, package.json, CLAUDE.md, and more
Stack detection across 200+ known packages in Python and JavaScript ecosystems
Prospect scoring 0-100 based on stack match, AI tool signals, production readiness, and activity
ATS job probing across Greenhouse, Lever, Ashby, and Workable — no API keys needed
Contact enrichment via Hunter.io and Apollo.io (BYOK model)
AI-powered outreach generation using resume + prospect data via Claude
Provider adapter pattern — external APIs abstracted behind mockable interfaces
132 tests passing across 13 test files
Business Impact
Finds companies building with your exact tech stack using public GitHub data
Eliminates ghost job applications by connecting directly to real engineering teams
Caches all external API responses (24hr-30day TTLs) to minimize rate limit consumption
BYOK model means free tier costs nothing to run
Technical Achievements
Provider adapter architecture that makes the system provider-agnostic
Auth design separating identity (Google OAuth) from data access (GitHub OAuth) from paid services (BYOK)
Caching layer that makes the database grow organically as users scan
132 tests with TDD as the standard
Technical Implementation
7 Django apps with clean separation of concerns. External APIs abstracted behind provider adapters — views and tasks receive provider instances; tests inject mocks. Identity separated from data access: Google OAuth for login, GitHub OAuth for API access, Hunter/Apollo as BYOK. All multi-step workflows run as Celery tasks with frontend polling via TanStack Query.